Daogao Liu

FlexOLMo: Open Language Models for Flexible Data Use

Private Geometric Median in Nearly-Linear Time

Adaptive Batch Size for Privately Finding Second-Order Stationary Points

Scaling Embedding Layers in Language Models

URANIA: Differentially Private Insights into AI Use

Improved Sample Complexity for Private Nonsmooth Nonconvex Optimization

Differentially Private Bilevel Optimization: Efficient Algorithms with Near-Optimal Rates

Unlearn and Burn: Adversarial Machine Unlearning Requests Destroy Model Accuracy

MUSE: Machine Unlearning Six-Way Evaluation for Language Models

Improved Sample Complexity for Private Nonsmooth Nonconvex Optimization

Scaling Embedding Layers in Language Models

Mind the Privacy Unit! User-Level Differential Privacy for Language Model Fine-Tuning

Private Online Learning via Lazy Algorithms

Detecting Pretraining Data from Large Language Models

Private Stochastic Convex Optimization with Heavy Tails: Near-Optimality from Simple Reductions

Faster Algorithms for User-Level Private Stochastic Convex Optimization